Mean cross-run high-band SOZ AUC change was 0.0031 lower for 30-second than 10-second samples in 34 SEEG patients

We tested sample length in 34 patients with public stereoelectroencephalography (SEEG) data from OpenNeuro ds004100, version 1.1.3. Each patient had two 60-second recordings. The fixed score used mean 80 to 150 Hz power and omitted 118 to 122 Hz. Larger power predicted the clinical seizure onset zone (SOZ) label. AUC measured the order of SOZ and other pair scores. The outcome was the mean absolute AUC difference across all interval combinations between recordings. Thirty-second samples had 0.003116 less mean variation than ten-second samples. The planned patient sign-flip test gave p = 0.06726. Evidence for a duration effect was uncertain. Using the labels to select the best ten-second interval raised AUC by 0.03682 compared with the first interval. This selection used the same data and does not estimate performance in new patients. All 68 raw prefixes and 1,795,500 power values passed source checks. A clean rerun reproduced all numerical files. The package preserves arrays, metadata, code, results, figures, and pinned raw retrieval instructions. The study does not establish a sufficient clinical recording duration.

stereoelectroencephalography · epilepsy · seizure onset zone · high-frequency power · sample duration · area under the curve · interval selection · repeatability